While what ATi did was indeed cheating in the context of 3DM03 being a test of generic DX8/9 games (without card-specific code), it's still an ethically superior cheat than nVidia's because it's conceivable that ATi's drivers will reach a point in the future where they can automatically shuffle shader code for optimal execution speed for all 3D apps. Drivers that can foresee where your camera will look next and add clip planes accordingly, OTOH, I don't see becoming available. As for lowering precision, that's not acceptable at all. Give a customer the choice for lower precision and faster performance, like with 32- and 16-bit color depths, but don't make the Folger's switch without notifying them. I'd be interested to see if NV3x cards will actually run at full precision when set to the ARB2 code path in Doom 3, or if they'll stay in the visually similar (but inferior) and faster-performing NV30 code path.

Both companies done bad, but nVidia done worse. Still, I can't say for sure ATi would have taken the high road if they were not the current clear DX9 leader. All I know is ATi currently offers the best cards from $150 to $400, in terms of both features and price per performance.

Droopy:

I don't think ATi cheated the same way Nvidia did.

It's like comparing a misdemeanor with a felony. That's what he meant.

You go in for an interview and they ask about your history. If you told them you jacked a dude up once, I think that's a lot better than you telling them you stole a few cars and killed some folks in the process.

1.9% is a whole lot less than 24%. Totally deleting rendered objects is a lot worse than just reordering instructions. A lot worse. In the end, Nvidia drastically changed the output, while ATi didn't.

Look at the frame comparison on Beyond3d. Nvidia had changed pixels a 10 fold more than ATi did.

This is the real issue here. Where do you draw the line? At what point is a company trying to screw us? Everyone seems in 100% agreement that what Nvidia did was wrong. They deserve all the bad press, PR, and general dislike for what they have done from the public.

What is interesting here is there seems to be two camps on what ATI did. Some people are like "They only broke the rules a little bit," while on the other side there are those who ar like "Rules are rules and they where broken." Clearly ATI would have been better served if they had not optimized that code. It is also clear by them coming out about it that ATI knows this.

What side of the camp you fall in is a personal thing. I tend to think that both of them where wrong, but what Nvidia did was much easier to condem as unethical. ATI is on the right track to earn my trust back in all of this. Nvidia on the other hand seems to be content to stay in the dog house witht he public.

A good "What if ?" is: "What if ATI had been the only company to be caught 'optimizing' their drivers for 3DMark03 ?" The Nvidiots would be howling for blood right now...

What ATi did was a detected game optimization with the exact same output, will Nvidia did a detected game cheat, with a reduction in quality and output.

Some Nvidiots can't see that, and just say a cheat is a cheat.

The way Nvidia brushes it off, by saying Futuremark isn't want they believe to be a valid benchmark. If that's the case then why did they optimize and cheat?

And it's really lame when Hard OCP does what Nvidia tells them to do.

Is there no honor among theives!!!!!!!!!!!!!

Now you have a terrible example Buub. The point is that what Ati did wouldn't be considered a cheat in a game. ATI PRESERVED THE SAME IQ!!!
A real good analogy would be that Ati was carrying a knife and Nvidia brought a rifle :lol:

in my opinion, ATI check Nvidia.

On the same page: http://www.beyond3d.com/#news6041

Comments from the creator of the Unreal engine on the difference between driver cheating and driver optimizations:

Pixel shaders are functions, taking textures, constants, and texture coordinates as inputs, and producing colors as outputs. Computer scientists have this notion of extensional equality that says, if you have two functions, and they return the same results for all combinations of parameters, then they represent the same function -- even if their implementaions differ, for example in instruction usage or performance.

Therefore, any code optimization performed on a function that does not change the resulting value of the function for any argument, is uncontroversially considered a valid optimization. Therefore, techniques such as instruction selection, instruction scheduling, dead code elimination, and load/store reordering are all acceptable. These techniques change the performance profile of the function, without affecting its extensional meaning.